Representation and Mimesis in Generative Art:...
Transcript of Representation and Mimesis in Generative Art:...
Representation and Mimesis in Generative Art: Creating Fifty Sisters
Jon McCormack
Centre for Electronic Media Art, Monash Univeristy, Caulfield East, Australia
Keywords: Generative Art, Representation, Mimesis, Artificial Life.
Abstract: Fifty Sisters is a generative artwork commissioned for the Ars Electronica
Museum in Linz. The work consists of fifty 1m ∞ 1m images of computer-synthesized
plant-forms, algorithmically ‘grown’ from computer code using artificial evolution and
generative grammars. Each plant-like form is derived from the primitive graphic ele-
ments of oil company logos. The title of the work refers to the original ‘Seven Sisters’ — a
cartel of seven oil companies that dominated the global petrochemical industry and
Middle East oil production from the mid–1940s until the oil crisis of the 1970s.
In this paper I discuss the issue of representation in generative art and how dialogues in
mimesis inform the production of a generative artwork, using Fifty Sisters as an example.
I also provide information on how these concepts translate into the technical and how
issues of representation necessarily pervade all computer-based generative art.
Computation Communication Aesthetics and X. Bergamo, Italy. xcoax.org
71
1. Introduction
In a recent paper, the author and several colleagues proposed what we considered to be
the ten most important questions for generative art (McCormack et al. 2012). The fifth
question on our list asked the following in relation to computational generative art:
In what sense is generative art representational, and what is it representing?
In this paper, I will expand on this question and its implications. From the onset I
should make clear that my topic relates to computational generative art. While generative
art has many non-computational modes, they are not specifically addressed in this paper.
Rather than discussing theoretical ideas, I will describe a recently completed generative
art work, Fifty Sisters, and look at how representational issues come into play in almost
every aspect of creating the work: conceptualization, implementation, and realization.
Representation and mimesis are some of the oldest issues in art, dating back at least to
the ancient Greek philosophers (Scruton 2009). The idea of replicating naturalistic effects
in painting came to the fore in renaissance aesthetics, where painters were concerned
with a truthful representation of what they saw. Roughly corresponding with the math-
ematical formalization of perspective projections and with progressive advances in paint
technologies (Ball 2002), artists’ skills developed in portraying the ‘real’ in art. However,
any art acting as a ‘mirror of nature’ — as was famously advocated by Leonardo — still
requires ‘interpretation and ordering’ from the artist. Image reproduction technologies
forever changed the idea of capturing the real in art, inviting the possibility for artists
to focus on other kinds of ‘truths’. By the time Hal Foster published Return of the Real
(Foster 1996) mimesis had come full circle1. In more recent times, representation and se-
miotics arguably have been overtaken by other concerns, such as art as social exchange
and dialogues concerning relational aesthetics. Representation and mimesis are old and
well established discourses in art. Generative art reopens these dialogues in ways that
other art forms cannot, because generative art brings something new to art: the idea of
representing process.
In generative art, as with other forms of art, we should expect a range of representa-
tional styles, e.g. visual art ranges from abstract, non-objective mark making and vast
sways of negative space, to highly figurative and photorealistic imagery. But unlike visual
art, generative art has not been so extensively analyzed in terms of how it deals with
mimesis and representation. Here we seek to begin to address that deficiency.
1.1. Computers and RepresentationIt is almost impossible to write a computer program without — at least implicitly — con-
sidering representation. Digital computers use collections of bits (electrical signals or
states standing for 1s and 0s, ON and OFF, etc.) to encode and represent data and instruc-
tions. At the level of software, programs generally represent things using atomic variables
(integers, floating-point numbers, Booleans, characters, etc.) or compound collections of
these variables (data structures, arrays, strings, objects), which may easily include other
compound collections.
1. In fact a circle and a half.
72
It is important to distinguish between a variable and its interpretation, i.e. it’s se-
mantics. I can give a variable any semantic interpretation I choose: it could represent
happiness, my bank balance, or the text of this paper for example. The important point is
that the programmer, not the computer, confers this meaning. The computer hardware
does impose practical limitations on the kinds of interpretations that are possible. If I
represent the concept of happiness as an integer, then the machine will manipulate it as
an integer, not as an emotional state of being. I can interpret a variable called ‘happiness’
as the degree of happiness of an individual, but this representation is limited to the 32 or
64 bits typically used to represent integers.
At the base level, computers are symbol-processing machines — they transform pat-
terns of bits that represent symbols. All symbols are subject to interpretation, either by
the programmer, the user of the program, or the machine itself. As an added complica-
tion, symbols are often not interpreted directly (as bit patterns), but are transformed by
some process.
As an example: two variables may represent the Cartesian coordinates of the center of
a circle that is displayed graphically on a computer screen. As the variables are changed
the circle ‘moves on the screen’. An additional variable represents the circle’s radius. We
speak of a ‘circle moving on the screen’, but, like the world of the Matrix, there is no cir-
cle and it is not moving. Discrete patterns of bits are changing at regular intervals. We
interpret the complex process of changing individual pixels seamlessly as a moving circle.
Perhaps this all seems obvious and even slightly trivial. Yet it is common for peo-
ple working with computers to forget about these representational gaps. In observing a
computer simulation of ant behavior, we might speak of ‘ants foraging for food’, but this
interpretation (as it would be for a painting) is semantically loaded. They are only “ants”
in as much as they homomorphically model an ant. ‘Foraging’ is a convenient anthropo-
morphic label we give to a series of discrete changes read as position, movement, behavior
and so on. We use this shorthand because it is both convenient and necessary: speaking
only in terms of bit patterns is not practical or enlightening (Dennet 1991), despite this
being the basis of digital computer representations.
As Nietzsche reminds us, writing on a typewriter is different than writing with a pen
(Kittler 1990). The tool affects our way of thinking. With the computer it is even more
profound, because to translate ideas into code we must ‘think algorithmically’, which in
turn influences how we think about the world and act in it.
In making a generative computer artwork, representation exists at many levels, not
just the bit-pattern level of variables, data structures or screen graphics. Rather than
expand on this in abstraction, let us look at a concrete example in generative art to see
how these issues come into play.
2. Fifty Sisters
Fifty Sisters is a generative artwork commissioned for the Ars Electronica museum in
Linz, Austria2. The work consists of fifty 1m x 1m digital images of computer-synthesized
plant-forms, arranged in a 5 x 10 grid in the museum foyer. Each image is algorithmical-
ly ‘grown’ from computer code using artificial evolution and generative developmental
grammars. The form is derived from the primitive graphic elements of oil company logos.
2. www.aec.at
73
The title of the work refers to the original ‘Seven Sisters’ — a cartel of seven oil compa-
nies that dominated the global petrochemical industry and Middle East oil production
from the mid–1940s until the oil crisis of the 1970s. Fossil fuels began as plants that over
millions of years were transformed by geological processes into the coal and oil that
currently powers modern civilization. The images remind the viewer that the basis of
an oil company’s financial success is derived from plants and natural processes that op-
erated over vast geological timescales. With ‘peek oil’ expected to be reached this century
(if not already), we are expending this non-renewable resource in the relative blink of
an eye. Two example images from Fifty Sisters are shown in Figure 1. More information
on the work and its motivations can be found at http://jonmccormack.info/~jonmc/sa/
artworks/fifty-sisters/.
The process to create each form involved a number of steps. Firstly, an oil company
logo was chosen and 2D vector art created3. From the 2D vector art, the basic graphic ele-
ments were separated manually and then converted to 3D geometric primitives. To cre-
ate each plant form custom software was developed by the artist. Technical details can
be found in (McCormack 2005). In basic terms, the software simulates the growth and
development of the form from a series of developmental rules, metaphorically similar to
the way DNA ‘encodes’ the developmental plans of biological organisms.
Fig. 1. Synthesized plant forms based on the BP logo (left) and ESSO logo (right).
Rules consist of any number of developmental symbols that represent individual or col-
lective elements of the growing form. Symbols include continuous data, such as size, age,
chemical concentrations, etc., that change over the lifetime of the developmental simu-
lation. If certain conditions are met (e.g. size becomes greater than some fixed value), the
symbol may subdivide, be replaced by another symbol, or die. This method is somewhat
analogous to cell division in biology, but with far greater abstraction and simplification.
As the developmental rules are a machine-representable code, they can be subject
to genetic manipulation, including mixing of rules from other forms (a kind of ‘gene
splicing’) or guided evolution using a variant of the Interactive Genetic Algorithm or IGA
(McCormack 2004). The terminal symbols of any rule can be interpreted as instructions
that encode the geometric construction of form. These symbols include instantiations of
3. Wikipedia has vector versions of many oil company logos.
74
the geometric elements of the oil company logos. Thus when the form is constructed, its
geometry includes geometric elements of the original logo. The final form depends on
how the rules have evolved and mutated. The results are often surprising; in some cases
the original logoform is clearly visible, in others it is almost impossible to recognize as
it has become highly abstracted. Figure 2 shows an example form (using elements from
the Shell logo) and the rules, or ‘digital DNA’ used to generate it.
Fig. 2. Synthesized plant forms (left) and the developmental rules from which it was generated (right).
The forms generated by this developmental/genetic process are output as 3D geomet-
ric models. Most plant forms are easily expressed in only a page or two of information
(a few hundred bytes), yet they generate geometric models many orders of magnitude
greater (~107 – 109 bytes). The models are read into a 3D renderer, which renders an image
using “photorealistic” rendering techniques.
2.1. Representation in Fifty SistersFifty Sisters is a useful example of representation in generative art, because it deals with
representation and mimesis at multiple levels. As images, each plant form has several
representations: that of a real plant, a computer graphic, and a corporate logotype. The
generative code (‘digital DNA’) from which each image is generated (Figure 2) is also ex-
hibited in a separate touch screen application that forms part of the exhibition of the
work. This allows the viewer to see a different representation of the form: as code that
through a process mimetic to biology generates that form.
Beyond the visual and textural representations, there is an additional layer of rep-
resentation to contend with, that of the generative process. What is represented in this
process? The process represents another process: biological development and evolution.
The artist-developed computer program simulates and abstracts the process of biological
development and evolution. It is the personal expression of a biological process in software.
So in this sense the software program, when run, represents these natural processes in
a somewhat similar way to that in which a landscape painting represents a landscape.
75
The difference, of course, is that the viewer of the work cannot see or otherwise experi-
ence this process directly.
This idea of one dynamic process representing another is new to art, and is what
best distinguishes this kind of generative art from other practices. Certainly process was
often of interest in modern art. One only has to think of Sol Le Witt, Cornelius Cardew
or Jackson Pollack for example. But in these cases the process of generating the art was
not representing another process: Le Witt’s drawing instructions were not representing
anything other than instructions to draw. A computer process being mimetic to another
process is different, because it involves choices about sign, signifier and what is signified.
Moreover the complexity and unpredictability of a computer process (vis. Emergence
(McCormack & Dorin 2001)) introduces additional properties not directly represented in
the generative process itself.
2.2. Mimesis in Fifty SistersAs alluded to in the introduction, mirroring nature involves interpretation and ordering
by the artist. As simulacra or simulation, a computer process is not the same as what it
seeks to mirror. This is well known in the simulation sciences, were formal methods are
used to verify and validate simulations to models, and models to reality. The experimenter
selects those aspects to model and those to ignore. Naturally, aspects or mechanisms that
the experimenter is unaware of cannot be in the model, although through experimenta-
tion she or he may become aware of them, and then subsequently incorporate them into
the model. The aspects of a phenomena or system that are modeled are subject to varying
degrees of abstraction necessary for them to be practically simulated.
Art allows a difference license, where the most interesting works can abstract from
the world of the imagination rather than the world of the real. In Fifty Sisters mimesis
plays many roles. The plants themselves are in some way mimetic to real plants, yet no
such forms could ever exist in reality. This is not really surprising; such issues have been
endlessly explored in painting.
Things become more interesting in relation to process however. The generative pro-
cess is mimetic to real biological development and evolution. The work speaks of ‘digital
DNA’, ‘evolution’ and ‘development’ as signifiers to the interpretation of their biological
parallels. While this simulated biology is grossly abstracted and simplified, it still exhibits
some of the features of its real-world counterpart. Moreover, its conceptualization and
intent as artistic concepts originates from interpretations of biological development and
evolutionary process.
This analysis reveals some curious aspects about the work. For example the choice of
using standard 3D rendering techniques, which focus on a Cartesian, photographic-like
visual realism, whereas the biological processes focus on a somewhat different kind of
‘realism’. This is partly explained by the technical constraints in developing works like
this, but more importantly the aesthetic language of modern corporate communication
is similarly derived from these techniques. Corporate logos are visually presented using
the purity of glittering computer graphics, with its clean and sleek mathematical veneer.
Fifty Sisters deliberately borrows from this vernacular, presenting developmentally man-
gled corporate logo forms using their native visual language.
76
3. Conclusion
In order to understand a generative artwork we must examine the process alongside
what that process produces. It is important to look at what the process is representing,
and how it performs this representation. For generative computer art, there always exist
multiple levels of representation and it is easy to forget about how these representational
structures are formed when they are so easily taken for granted. One of art’s roles can
be to reveal what is normally hidden or taken for granted, bringing it into awareness (or
even sub-consciousness). Computer representations and processes are typically hidden
from direct perception, so by bringing them into perception we reveal their most unique
and interesting aspects as symbol processing machines.
# 1 “bpOldZoom.dna”
# 1 “<built-in>”
# 1 “<command-line>”
77
object spine {
<< noises >> efn;
<< math >> efm;
< efn.gauss > gauss;
< efm.acos > arccos;
< efm.sin > sin;
surface BP_old_hood;
surface BP_old_back;
surface BP_old_BP;
surface sphere10;
equiv ‘ col;
rules:
el(i) : i <= 47 -> /(137.5) [ ^(arccos(1 - i * 0.04000) )
[ f(90) elem(i) ]] el(i+1);
elem(i) -> sph seg(100,40,5,1,10,0.1,0.1);
sph -> !(15) c(0) !(7) C(1,0) !(3) C(1,0) !(1) C(1,0);
seg(n,l,t,u,r,sc,v) : n > 0 -> /(t) ^(u) !(r * sc) C(l,0) [col(1)
S(sc) BP_old_hood ] seg(n - 1, l + gauss(0,v), t + gauss(0,v), u
+ gauss(0,v), r, sc + gauss(0,v), v)
: n <= 0 -> /(t) ^(u) !(r * sc) C(l, 1) [ col(9)
tusks(36)
] f(14) S(1.0) ll(18, 15, 10);
tusks(n) : n > 0 -> [ ^(90) !(14) c(0) C(100, 0)
tusk_s(20,gauss(90,5),14,5)
] /(10) tusks(n - 1);
tusk_s(n,l,r,u) : n > 0 -> !(r) /(gauss(0,5)) &(u) C(l,0) tusk_s(n
- 1, l *
0.90, r * 0.78, u * gauss(1.08, 0.067))
: n <= 0 -> !(r * 0.1) ^(u) C(l,0);
ll(i,u,t) : i > 0 -> /(20) [ &(u) +(t) col(2) BP_old_hood col(1)
BP_old_back
] ll(i - 1, u, t)
: i <= 0 -> [ f(5.0) col(1) lm(18,u+10,t - 4.5) ];
lm(i,u,t) : i > 0 -> /(20) [ &(u) +(t) S(0.5) col(3) BP_old_hood ]
lm(i - 1,
u, t)
: i <= 0 -> [ f(5.0) col(1) ls(6,u + 10,t - 10) ];
78
ls(i,u,t) : i > 0 -> /(20 * 3) [ &(u) +(t) S(0.25) col(3) BP_old_
hood col(1)
BP_old_back col(1) BP_old_BP ] ls(i - 1, u, t);
axiom:
@(0.2) *(2,0,1) col(0) [col(8) S(9) sphere10] el(0);
}
scene {
spine(time * 50);
}
Fig. 3. Synthesized plant form based on the BP logo (top) and the developmental rules from which it was generated (bottom).
Acknowledgements: I am grateful for discussion with Gordon Monro who first raised the
question of representation in generative art with me. This research was supported by an
Australian Research Council Discovery Grant, DP1094064.
References
Ball, P. Bright earth: art and the invention of color. New York, Farrar Straus
and Giroux. 2002.
Dennett, D. C. Real Patterns. Journal of Philosophy 88: 27–51. 1991.
Foster, H. The return of the real: the avant-garde at the end of the century. MIT Press,
Cambridge, Mass.1996.
Kittler, F. Discourse Networks 1800/1900, with a Foreword by David E. Wellbery.
Stanford. 1990.
McCormack, J. Aesthetic Evolution of L-systems Revisited. Applications of Evolutionary
Computing (EvoWorkshops 2004). G. R. Raidl, S. Cagnnoni, J. Brankeet al. Berlin,
Heidelberg, Springer-Verlag. LNCS 3005: 477–488. 2004.
McCormack, J. A Developmental Model for Generative Media. Advances in Artificial
Life (8th European Conference, ECAL 2005). M. Capcarrere, A. A. Freitas, P. J.
Bentley, C. G. Johnson and J. Timmis (eds). Berlin; Heidelberg, Springer-Verlag.
LNAI 3630: 88–97. 2005.
McCormack, J., O. Bown, A. Dorin, J. McCabe, G. Monro and M. Whitelaw. Ten
Questions Concerning Generative Computer Art. Leonardo (to appear, accepted July
2012).[preprint available at: http://www.csse.monash.edu.au/~jonmc/research/
Papers/TenQuestionsLJ-Preprint.pdf] 2012.
McCormack, J. and A. Dorin. Art, Emergence and the Computational Sublime. Second
Iteration: a conference on generative systems in the electronic arts, Melbourne,
Australia, CEMA. 2001.
Scruton, R. Beauty. Oxford University Press, Oxford; New York. 2009.
79